Model Details
Mistral Small 4
Quick answer: Mistral Small 4 is Mistral AI's September 2025 Small tier model scoring 78% MMLU-Pro and 58.3% Arena Hard. Proprietary — Mistral Small generation 4.
At a Glance
Where Mistral Small 4 leads
- 78% MMLU-Pro — improved over Small 3 (66.3%)
- Mistral API availability
Where it lags
- 58.3% Arena Hard — well below Small 3 (87.6%)
- Proprietary license
Best for: MMLU-Pro knowledge tasks via Mistral API; Small-tier budget.
What Mistral Small 4 Is
Mistral Small 4 is the September 2025 update to the Mistral Small series. MMLU-Pro jumps from 66.3% (Small 3) to 78%, while Arena Hard drops from 87.6% to 58.3%.
Specifications
| Field | Value |
|---|---|
| Organization | Mistral AI |
| License | Mistral Small 4 License |
| Release date | September 2025 |
| Modality | Text only |
Pricing
Available via Mistral AI API.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| MMLU-Pro | 78% | Benchgen evaluation | 2025-09 |
| Arena Hard | 58.3% | Benchgen evaluation | 2025-09 |
Mistral Small 4 vs Alternatives
| Model | MMLU-Pro | Arena Hard | License |
|---|---|---|---|
| Mistral Small 4 | 78% | 58.3% | Proprietary |
| Mistral Small 3.2 24B | — | — | Apache 2.0 |
Mistral Small 3 (87.6% Arena Hard, Apache 2.0) likely better for instruction tasks. Small 4 only for MMLU-Pro with latest API.
Related Models & Benchmarks
- Compare: Mistral Small 3.2 24B · Mistral Small 3.1 24B
- Benchmarks: MMLU-Pro · Arena Hard
Frequently Asked Questions
What is Mistral Small 4?
Mistral AI's September 2025 Small tier scoring 78% MMLU-Pro, 58.3% Arena Hard.Specs from Mistral AI's Mistral Small 4 release (September 2025) and Benchgen evaluations. Last updated 2026-07-24.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.